Optimal Motion Planning for Object Picking in Industrial Contexts with Optimal Control
Dries Dirckx, Jan Swevers, Wilm Decre
AI summary
Problem
Industrial manufacturing requires motion planners that adapt quickly to variable configurations, but existing GPU-based optimizers often rely on conservative approximations that compromise accuracy and safety, while lacking hardware accessibility in typical assembly cells.
Approach
The method formulates pick-and-place motion as a hard-constrained time-optimal optimal control problem solved on standard CPU hardware, enhanced by a near-optimal warm-starting strategy that reuses previously computed trajectories for similar tasks.
Key results
- 98.33% success rate compared to 81.67% for cuRobo
- Guaranteed final pose accuracy within 1 mm and 1°
- 0.70× lower execution cycle time than GPU-based planners
- Up to 57.5% computation time reduction via near-optimal warm-starting
Why it matters
It offers a reliable, hardware-accessible alternative for flexible manufacturing cells that require precise, safe, and time-optimal robotic motion without dedicated GPU infrastructure.
Abstract
No abstract on file.